I'm curious whether anyone out there performs stress tests on their new servers—specifically regarding CPU and RAM—before rolling them out. Or do most people just trust that everything will be fine, deploy the servers, and let support handle any issues that may arise? I'm looking at a Dell R360 and wondering what's the best practice.
5 Answers
Stress testing feels outdated to me. About 20 years ago, I did do a long run on new servers just to configure them slowly, but nowadays, I think it’s unnecessary. Those who have deployed a ton of servers might disagree, but I believe in sending it out once it’s built.
In my experience, we usually skip the stress testing on factory-built servers. The manufacturers typically handle that part. As long as the server powers on and shows green status, we feel confident deploying it.
I always run memtest64 on new servers. Skipping this once led to data corruption issues for me, so I really think it’s worth the time. For white box servers this helps, but prebuilt ones are typically more reliable right out of the box.
We don't stress test anymore. We just do a base install of VMware and run a canary VM to shake down the configuration. It’s less about stress testing and more about catching issues up front. Recently, we had a bad BIOS update that we caught early in the process, but outside of that, things have been smooth sailing!
We have a pretty laid-back approach: we install ESX, leave it idle for a couple of days, and then move some non-critical VMs over to test. I usually create a few dummy VMs and test storage, networking, and backups before switching anything important to the new server.
That sounds like a smart way to ensure everything's in order before going live. I'll consider that!
That's a good approach! Early detection saves a lot of hassle later on.